pi-config workflow

Personal Pi agent setup with four commands for plan, todos, execution and review, passing state through Markdown session files.

  • 459 ★ on the repository
  • Last commit
  • MIT
Workflow

Broad access

Why this level

  • `setup.sh` runs on your machine, installs Git-backed packages and on macOS a runtime that drives the desktop.
  • The `execute_command` extension lets the agent invoke slash commands itself, and workers create commits without asking.
  • The configuration replaces the global agent directory `~/.pi/agent/`, including models and `mcp.json`.
All reasons and checks

hazat/pi-config

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add workflows/pi-config

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • `setup.sh` runs on your machine, installs Git-backed packages and on macOS a runtime that drives the desktop.
  • The `execute_command` extension lets the agent invoke slash commands itself, and workers create commits without asking.
  • The configuration replaces the global agent directory `~/.pi/agent/`, including models and `mcp.json`.

This is third-party code. Review the repository files before installing.

What it does

The repository is cloned into `~/.pi/agent/` as a global Pi agent setup. It holds settings, models, `mcp.json`, prompt templates and skills. The working loop has four commands: `/plan` investigates the task and records an approach, `/todos` splits the plan into worker-ready tasks, `/execute` runs workers one by one and verifies each commit, and `/review` assembles an independent review team and repairs findings. State between steps lives in plain Markdown files beside the session file: `.plan.md`, `.todos.md` and `.review.md`. The `sc` (Superconductor) CLI drives orchestration, and the `commit`, `scout`, `worker` and `review` skills set rules for individual roles.

Who it is for. Experienced Pi users who want a ready plan, todos, execute, review loop and are willing to adapt it.

Good fit when

  • A large task needs a plan and verifiable steps.
  • Workers should run in sequence and leave one commit each.
  • Finished work should be checked by an independent team, not the same agent.
  • A sample is needed for how Pi skills and prompts are organized.

Not a fit when

  • The agent is not Pi; the setup targets its directory and skill format.
  • `sc` is unavailable and orchestration does not work without it.
  • A neutral setup is wanted: the repository reflects one author's tastes.

Example request

/plan add CSV report export. Then /todos, /execute and /review against the finished plan.

Limitations

This is the author's personal setup, not a universal package. `setup.sh` requires the clone to live exactly at `~/.pi/agent/`, preserves an existing `settings.json`, installs Git-backed packages and on macOS adds the macOS Harness runtime through `uv`. It needs `PI_SESSION_FILE`, provider keys in `auth.json` and the `sc` CLI. README verification relies on `jq` and the `gpt-5.6` model. Skills and prompts reflect the author's tastes, so read and adjust them for your own team before use.

How to disable. Stop Pi and restore your earlier `~/.pi/agent/` from backup, or delete the clone. Packages installed by `setup.sh` and the macOS Harness runtime are removed separately.

FAQ

What does `/execute` do?

It reads the plan and todos, runs workers in sequence, verifies each commit through Git and updates the todos file from the coordinator session. Workers never write to the todos file.

How many repair rounds does `/review` run?

At most two. On a `NEEDS CHANGES` verdict sequential workers run, then a fresh review repeats until approval or a bounded stop.

Are `/plan` and `/todos` required before `/review`?

No. Review needs only the session JSONL, and the plan and todos files are optional context.

Where does Git work happen?

The workflow does not create branches, push or open PRs. Those steps happen only on a separate explicit request.

Which skills ship in the repository?

`plan`, `write-todos`, `worker`, `review`, `scout`, `superconductor`, `commit` and others. The full list is in the directory layout table in the README.

Discussion

No questions or reviews yet.

Workflows
Editors’ pick

A minimal terminal coding agent extended with skills, prompt templates and TypeScript packages

CLIBroad access121.8KRepository stars

A set of extensions for the Pi agent: repository analysis, secrets leased with a TTL and search over agent sessions.

PluginExtended access87Repository stars

Background runs of pi coding agents on your machine by schedule and label, each job in a Docker container with a spending limit.

CLIBroad access197Repository stars

A web app for the Pi coding agent: sessions survive restarts, teammates can join, and you can work from a phone with no cloud VM

CLIBroad access833Repository stars

A web dashboard for the pi coding agent with sessions grouped by branch, a built-in terminal and a mobile mode

CLIBroad access348Repository stars

Repeated pi agent steps accumulate as JavaScript programs with a manifest and tests, and pi calls them instead of redoing work.

PluginExtended access297Repository stars

README in short

The README covers installation into `~/.pi/agent/`, the four loop commands, the Superconductor policy and the directory layout. State is kept in Markdown with no sidecar stores or run directories.

README badge

Report a problem